Académique Documents
Professionnel Documents
Culture Documents
EB-3105 PAGE 2 OF 13
Case Study
ROI for a Customer Relationship Management
Initiative at GST
Mary Gros
CEO
Tom Webster,
COO
Mark Johnson, Jeff Richards, Daniel Wymer, Barb Young, Cheik Daddah, Erica Kolks, Nichole Knell, Karine Hatti,
CFO CIO CAO General Counsel Investor Relations Marketing Industry Relations Human Resources
Stacy Hoyle, Jill Newburg, Dominique Arnold, Jeff Shoemacher, Meghan McCormick, Raveen Rajavama, Jean Secrist,
VP Region #1 VP Region #2 VP Region #3, VP Region #4 VP Region #5 VP Region #6 VP Region #7
1a
Jeff Shoemacher,
CEO,
VP Region #4
Fall Ainina, Rebecca Koop, Susan Lightle, Bud Baker, Paula Saunders, Cathy Kempf, Michael Edwards, Joe Castellano,
CFO CIO CAO ILEC CLEC Internet Services Data Services Customer Relations
1b
As the business evolved technologically The organization of each GST geographic TERADATA
and geographically, GST adopted a region includes a regional vice president serv- Teradata is a division of NCR Corporation, and
decentralized model by region. The ing as the CEO of the business unit, a regional is a leading provider of enterprise data ware-
corporate level leadership team includes CFO, a regional CIO who also reports to the housing technology and solutions.
the President and CEO, the COO, and corporate CIO, a CAO and several product NCR has a storied history dating back
fifteen vice presidents; seven are regional managers. Exhibit 1b represents the organiza- to its inception in 1884. In that year, John
VPs while the other eight include the Chief tion chart for GST Region 4. H. Patterson purchased the National
Financial Officer, Chief Accounting Officer, Manufacturing Company, maker of the first
Chief Information Officer, Senior VP Mary Gros, CEO, had requested a set of income mechanical cash registers, and renamed it
for Investor Relations, VP for Human statements reporting MIS expenses separate National Cash Register Company.
Resources, VP for Marketing, VP for Industry from Cost of Goods Sold. She noted the
Relations, and the General Counsel. The high- increase in IT expense each year, both in Extending from mechanical cash registers,
level corporate organization chart is provided dollars terms and as a percentage of revenue, NCR evolved into an innovative supplier
in Exhibit 1a. and charged Johnson with finding ways to of advanced Point of Sale Solutions, the world-
cut costs. Exhibit 2 contains the comparative wide leader in sales and shipment of
income statements for the past three years Automated Teller Machines (ATMs), and data
for Region 4.
EB-3105 PAGE 3 OF 13
Case Study
ROI for a Customer Relationship Management
Initiative at GST
EB-3105 PAGE 4 OF 13
Case Study
ROI for a Customer Relationship Management
Initiative at GST
Scalability A schematic diagram of a typical data ware- too costly and as a result, data marts
Scalability is the ability to support more house is shown in Exhibit 4a. The typical flow have proliferated. Data marts are smaller
users over time. For an EDW, scalability has of data to information is as follows: operational repositories of information that are for a
multiple dimensions: hardware, support data is generated through customer transac- specific business unit or process. Exhibit 4b
of user connectivity, and from a database tions. Data is then transformed into a consis- is a schematic of a company similar to
perspective the ability to support ever tent format into storage for later use. The GST that does not have a centralized data
increasing expectations for complex as well appropriate information is extracted and warehouse, but instead has a series of
as ad hoc query performance. Demands imported for summarization. The summariza- isolated data marts.
on a data warehouse increase exponentially tion might involve comparing sales across
as data and user volumes grow, update time, across products, and by profit margins. As independent systems, data marts are
frequencies increase, and the operational Similarly, data can be summarized by cus- often considered less expensive to operate.
feeder systems multiply. The Teradata tomer, across time, and across products by This is only true if one ignores many of the
solution has demonstrated scalability. profit margin. Finally, the summarized data is hidden costs associated with data marts. In a
presented as information for use in future 2001 Gartner report, it was determined that
Support for High User Concurrency business decisions. data marts were 70% more expensive to
One of the sure signs of a successful data operative per subject area than a comparable
warehouse is when more and more business The storage component of the data flow is data warehouse.
users want access to it. In some environments, the subject of data warehousing. In most
this demand presents a dilemma: Do you decentralized business environments, Data marts are usually constructed for an
accept all users and suffer performance data warehouses have been considered individual user/business unit because of the
degradation that leads to diminished ware-
house effectiveness and user attrition? Or do
you restrict data warehouse access to a limit- Data Mart and Data Warehouse Architectures
Data
BACKGROUND ON Transformation
EB-3105 PAGE 5 OF 13
Case Study
ROI for a Customer Relationship Management
Initiative at GST
Type 2,
Data integration is essential to the development ODS Warehouse Data Mart
EB-3105 PAGE 6 OF 13
Case Study
ROI for a Customer Relationship Management
Initiative at GST
Enterprise,
Warehouse &,
Management
BACKGROUND ON
DATA MART SYSTEMS
2
Note that marketing research studies may provide additional insights into what constitutes
a reasonable percentage increase in value.
EB-3105 PAGE 7 OF 13 3
This is where some thought will have to be given as to what marketing actions will be taken.
Case Study
ROI for a Customer Relationship Management
Initiative at GST
AWS
data redundancy leads to staff redundancy, and BYNET BYNET BYNET BYNET
Up to ,
256,
eliminating disparate data marts can reduce BYNET BYNET BYNET BYNET
Cabinets,
,
the staff count. SMP SMP SMP SMP
Up to,
2,048,
SMP SMP SMP SMP
Intel CPUs
shown schematically in Exhibit 5a. The bottom UPS UPS UPS UPS
The Teradata EDW architecture consists of 2–512 processing nodes (each node consists of four high performing Intel based CPUs—this is
called a symmetric multi processor (SMP) node with disk scalability up to 100s of terabytes via highly available, hot-pluggable, Redundant
The middle section of Exhibit 5a contains node Array of Independent Disks (RAID) for data storage. Nodes can be aded in pairs to map to the processing requirements of each configuration.
Disk options exist with Teradata sourcing RAID configurations from EMC2 and LSI Logic. The GST data mart consolidation pilot system
would be approximately 20% of the complete EDW, and is shown schematically in the dashed box.
cabinets. Each cabinet has two nodes comprised
5a
of 4-Intel processors. In addition, nodes are
interconnected via Teradata’s BYNET. The processing
cabinets are designed for resiliency with uninter-
rupted power supply units in each cabinet. Up to
The Teradata BYNET
256 cabinets (equaling 512 nodes) can be configured
as a single massively parallel processing (MPP)
system.As of January 2002, a total of 2,048 Intel CPUs BYNET
could be configured in a complete Teradata EDW.
NODE, Cliques,
The top portion of Exhibit 5a presents the adminis- (4) CPUs in a Node,
Up to 512 Nodes
Grouping of 4 Nodes,
Redundancy in case of,
Node failure
tration work station (AWS). This is a standalone
UNIX or Windows based workstation that is the
RAID,
primary operations interface for MPP systems. Redundant Array of Indepent Disks,
Terabytes of Data
The AWS provides a single, graphical view of the
system. Not shown are the thousands of end users The BYNET is a high-speed interconnect that is optimized for parallel processing with the Teradata Relational
Database Management System. More specifically, two BYNETs are configured with every Teradata MPP
with access to the system. Finally, the dotted line (Massively Parallel Processing) System for redundancy, high performance and scalability. These BYNETs
are uniquely designed to provide simultaneous, bi-directional traffic (messages) between the :
¥ Processing Nodes
containing three cabinets (or six nodes) is the ¥ Parsing Engine (PE-checks the SQL statement, access rights and invokes action),
¥ and the Access Module Processors (AMPs) for effective data retrieval and disk management.
footprint for the proposed GST pilot program. The BYNET is the key design feature that enables support for many concurrent users and maximum
system throughput.
EB-3105 PAGE 8 OF 13
Case Study
ROI for a Customer Relationship Management
Initiative at GST
connect across the message passing layer – this Davis pointed out that with just the three and could be eliminated through data mart
layer is also known as the system bus or as the systems and four access points represented, consolidation potentially saving millions
BYNET. Exhibit 5b also shows nodes sharing a there are 14 redundant processes. For example, in IT expenses. In addition to the expense
common set of disk arrays grouped into what are the “Acquire” step is the bridge between an of redundant systems, there are expenses asso-
known as cliques. The clique grouping provides access point (customer or supplier) and the ciated with the loss of accuracy from any
for data redundancy in case of node failure. firm. In Exhibit 6a, the “Acquire” step in System inconsistencies in the way the data is stored
C is completely redundant. That is, the four and reported across the systems. A centralized
access points have been completely sampled data warehouse eliminates these expenses as well.
DATA MART CONSOLIDATION by systems A and B by the time System C is
PROJECT building its database. To support his position, Davis also created
a revised organization chart for the same
GST is operating fifty disparate data marts. The Company wide, with 50 disparate data marts, region in a data warehouse environment -
manufacturers of the data marts include Oracle, GST had a massive amount of redundancy. Exhibit 6b. In Exhibit 6b, the data sits on top
IBM, Informix, and Sybase. Davis suggested the This redundancy was expensive, unnecessary of the organization giving everyone immediate
consolidation of the data marts into an enter-
prise data warehouse (EDW) for two reasons.
A. Region 4 in the Data Mart Environment
are more readily quantified, the value of the Clean Clean Clean Clean Clean Clean Clean Clean Clean
for consolidation: four Oracle 8i systems and B. Region 4 in the Data Warehouse Environment
EB-3105 PAGE 9 OF 13
Case Study
ROI for a Customer Relationship Management
Initiative at GST
An IBM data mart required one Non-personnel support $1,000,000/yr $1,800,000/yr $125,000/month after the data marts
costs are decommissioned
system administrator, three data
8
base analysts, two ETL programmers,
three query programmers, one network number and type of GST employees required for reductions. The most likely scenario for
administrator, and two people working as each Oracle and IBM data mart, see Exhibit 8. staffing the proposed enterprise data
support staff. Non-personnel support costs warehouse is one system administrator, eight
for the IBM system was $1,800,000 per year. data base analysts, four ETL programmers,
Maintenance and upgrades for the IBM mart COSTS FOR THE ten query programmers, and three individuals
total $110,000 per year. TERADATA SOLUTION serving as support staff. Exhibit 8 also
summarizes the best, worst, and expected
Lightle gave Davis GST employee salary and The staffing requirement for the Teradata case scenarios for staffing the new Teradata
benefits information, see Exhibit 7. She also system depends, in part, on how GST system. The exact probabilities for the
gave Davis a summary breakdown of the management decides to handle the personnel GST staffing changes were not known,
EB-3105 PAGE 10 OF 13
Case Study
ROI for a Customer Relationship Management
Initiative at GST
The list prices associated with the acquisition Software $90,000 $190,000 $190,000 $190,000 $500,000
of the data warehouse are included in Exhibit Training and Professional Services Costs for the Teradata Solution
9. The consolidation of the five data marts will Expense Year 1 Year 2 Year 3
require five nodes. Although the first four Training See Exhibit 11: $15,000 per month -$0-
nodes are sold as individual units, nodes $15,000 per month
starting in May
beyond the fourth are only sold in pairs. The
Consulting See Exhibit 11: $125,000 per month $125,000 per month
prices quoted in Exhibit 9 are per node. The $125,000 per month
proposed system (nodes, software, and disks) after implementation
would be depreciated using the MACRS 5-year Data Storage Disk Costs
class life schedule assuming the mid-year $650,000 (For 2.8TBytes of data)
9
convention. The total cost for disks is estimated
Adapted from Steven Weber, Pricing Director, Teradata, a division of NCR
as $650,000. Maintenance/upgrades for the
nodes and software is 10% of the list price. ments of the total enterprise (including the
Professional services costs (business requirements associated with the remaining
Finally, the first year non-personnel support consulting) for the three years of the pilot 45 data marts). In addition, Davis was
costs for the Teradata warehouse, once the study are quoted at $125,000 per month suggesting GST begin work on the development
system is operational, is projected to be once the implementation project is complete. of customer relationship management programs
$1,500,000 (paid in monthly installments.) Exhibit 11 gives the detailed break down of the that would be possible with the more complete
professional service costs during the estimated view of the customer. The professional services
On behalf of Teradata, Davis can offer an 12-month implementation schedule. costs in years 2 and 3 were associated with
installed price for nodes and software at 30% the design of the data warehouse under a full-
Consulting costs decline dramatically after
off the list price. In addition, Teradata is willing consolidation EDW scenario and for the
to provide a $400,000 equipment credit against the first year because GST was being urged to development of CRM programs.
the purchase price if GST commits to the purchase the hardware and re-architect the
consolidation pilot study. However, the disks data structure at the beginning of the process, Training costs, separate from business
for the data storage would not be eligible for which front-loads the consulting fees. This was consulting, would be $15,000 per month for
the 30% discount. an alternative to acquiring a node, migrating the first two years – see Exhibit 11 for the start
the data, and re-architecting the system date of the training. Some of these costs were
sequentially. Davis was convinced the related to training the existing employees on
Summary GST financial assumptions former was in the best long-term interest the new system as well as training dislocated
Required return for project investments 14% of GST. Davis was also encouraging GST existing employees for other internal positions.
Corporate Tax Rate 38% to engage Teradata’s team of consultants Training would commence once the data marts
Inflation Rate: Non-personnel costs 5% to commence work on the development are loaded into the warehouse. For this ROI
Inflation Rate: Personnel costs 4% of a logical data model to address a analysis, the training costs would be expensed
10 holistic look at the information require- as incurred.
EB-3105 PAGE 11 OF 13
Case Study
ROI for a Customer Relationship Management
Initiative at GST
IMPLEMENTATION
PROJECT
Data Mart Consolidation Project Baseline
Exhibit 11 is a high-level schematic 1st Quarter 2nd Quarter 3rd Quarter 4th Quarter 1st Quarter
of the proposed data mart consolida- Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb
EB-3105 PAGE 12 OF 13
Case Study
ROI for a Customer Relationship Management
Initiative at GST
base-line the professional service implemen- of 38%, expected an inflation rate for non- ANALYSIS
tation cost would increase or decrease by personnel support costs of 5% annually, and
approximately $270,000. Davis had experi- expects salaries to increase 4% per year Following are some questions to consider with
enced 9 similar data mart consolidation across-the-board. In addition, GST was your analysis:
projects. Of these, 2 had come in under time considering retaining one Oracle mart for • What is the project ROI and the pay
at 10 months, 3 had taken 12 months, and 4 an internal training program. These data back period?
projects had run over to the full 14 months. are summarized in Exhibit 10. • Of the best, worst, and expected case
© 2002 by Mark Jeffery. No part of this publication may be reproduced, stored in a retrieval system, used in a spreadsheet, or transmitted in any for by means -
electronic, mechanical, photocopying, or otherwise - without the permission of Mark Jeffery. Teradata is a registered trademark and WorldMark is a trademark of
NCR Corporation. All other brand and product names appearing in this release are registered trademarks or trademarks of their respective holders. NCR continually
improves products as new technologies and components become available. NCR therefore, reserves the right to change specifications without prior notice. All fea-
tures, functions and operations described herein may not be marketed in all parts of the world. Consult your NCR representative for further information.
© 2002 NCR Corporation Dayton, OH U.S.A. Produced in U.S.A. All rights reserved.
www.teradata.com www.kellogg.nwu.edu
EB-3105 PAGE 13 OF 13